also see fdo#46757

a) We need to ignore redline-deleted text, but count redline-added text
b) each block of text is denoted by its end position in the model
   and where that maps to in the view so a hidden portion
   should record its end point not its starting point, and a non-hidden
   deleted portion should always record its end point
c) when mapping a model position to the view we take the offset of
   the model pos arg from the block end and use that to offset the
   mapped block-end view pos to get the final view pos. But for
   hidden portions that won't make a whole lot of sense, and
   end up offsetting into prior portions, so map all positions within a
   hidden portion to the same block-end view pos

add regression tests for these cases

Change-Id: I45c76bba47fd430bc3bccb5f919502660d415d9e